[Qemu-devel] [PATCH v6 0/1] Add support for Xen access to vmport

[Qemu-devel] [PATCH v6 0/1] Add support for Xen access to vmport

[Don Slutz]

Changes v5 to v6:
  Paul Durrant
    Do we need a forward declaration?
      Nope. Drooped.
    Added Reviewed-by.

Changes v4 to v5:
  Paul Durrant
    vmware_ioreq_t struct is not really a request any more. Maybe
    vmware_regs_t?
      Renamed various parts from vmware_ioreq to vmware_regs.  Also
      HVM_PARAM_VMPORT_IOREQ_PFN to HVM_PARAM_VMPORT_REGS_PFN.
    cpu_by_ioreq_id name implies the array is indexed by an id
    carries in the ioreq.
      Renamed cpu_by_ioreq_id to cpu_by_vcpu_id.
    Is cpu_get_vmport_ioreq_from_shared_memory worth its own
    function?
      Moved in-line.
    I don't think you need the barrier anyway.
      Dropped the barrier.
    Oh, I now realize you mean the same theoretical rather than
    actual limit, in which case this can be a build time check
    anyway.
      Switch to build time check, move to a better place.
    You could avoid passing state to both of them by setting
    current_cpu here couldn't you?
      Yes, moved state usage to handle_vmport_ioreq().

  Stefano Stabellini
    Error out if it fails with error != -ENOSYS.
      Done.

Changes RFC-v2x to v4:
  Stefano Stabellini
    Please try to get rid of the #ifdefs.
      Moved 2 #ifdefs into hw/xen/xen_common.h

Changes v2 to RFC-v2x:
  Paul Durrant
    Use a 2nd shared page.
      Added HVM_PARAM_VMPORT_IOREQ_PFN usage.

Changes v1 to v2:
   More info in commit message.

  Stefano Stabellini
    the registers being passes explicitely by Xen rather than
    "hiding" them into other ioreq fields.
       Added vmware_ioreq_t
  Paolo Bonzini & Alexander Graf
    Fixup env access
      Added cpu_by_ioreq_id.
      Set current_cpu in regs_to_cpu(), clear in regs_from_cpu().
      Drop all changes to vmport.c

Note: to use this with Xen either a version of:

[Qemu-devel] [PATCH] -machine vmport=off: Allow disabling of VMWare ioport emulation

or

>From f70663d9fb86914144ba340b6186cb1e67ac6eec Mon Sep 17 00:00:00 2001
From: Don Slutz <dslutz@verizon.com>
Date: Fri, 26 Sep 2014 08:11:39 -0400
Subject: [PATCH 1/2] hack: force enable vmport

Signed-off-by: Don Slutz <dslutz@verizon.com>
---
 hw/i386/pc_piix.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/hw/i386/pc_piix.c b/hw/i386/pc_piix.c
index 103d756..b76dfbc 100644
--- a/hw/i386/pc_piix.c
+++ b/hw/i386/pc_piix.c
@@ -234,7 +234,7 @@ static void pc_init1(MachineState *machine,
     pc_vga_init(isa_bus, pci_enabled ? pci_bus : NULL);
 
     /* init basic PC hardware */
-    pc_basic_device_init(isa_bus, gsi, &rtc_state, &floppy, xen_enabled(),
+    pc_basic_device_init(isa_bus, gsi, &rtc_state, &floppy, false,
         0x4);
 
     pc_nic_init(isa_bus, pci_bus);
-- 
1.8.4

needs to be done to QEMU.

And the Xen RFC patch:

[RFC][PATCH v2 1/1] Add IOREQ_TYPE_VMWARE_PORT

needs to be done to Xen.

Don Slutz (1):
  xen-hvm.c: Add support for Xen access to vmport

 include/hw/xen/xen_common.h |  22 +++++++++
 xen-hvm.c                   | 108 ++++++++++++++++++++++++++++++++++++++++++--
 2 files changed, 125 insertions(+), 5 deletions(-)

-- 
1.8.4

[Qemu-devel] [PATCH v6 1/1] xen-hvm.c: Add support for Xen access to vmport

[Don Slutz]

This adds synchronisation of the 6 vcpu registers (only 32bits of
them) that vmport.c needs between Xen and QEMU.

This is to avoid a 2nd and 3rd exchange between QEMU and Xen to
fetch and put these 6 vcpu registers used by the code in vmport.c
and vmmouse.c

The registers are passed in the new shared page provided by
HVM_PARAM_VMPORT_REGS_PFN.

Add new array to XenIOState that allows selection of current_cpu by
vcpu id.

Now pass XenIOState to handle_ioreq().

Add new routines regs_to_cpu(), regs_from_cpu(), and
handle_vmport_ioreq().

Signed-off-by: Don Slutz <dslutz@verizon.com>
Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
---
v6:
    Do we need a forward declaration?
      Nope. Drooped.
    Added Reviewed-by from Paul Durrant.

v5:
    vmware_ioreq_t struct is not really a request any more. Maybe
    vmware_regs_t?
      Renamed various parts from vmware_ioreq to vmware_regs.  Also
      HVM_PARAM_VMPORT_IOREQ_PFN to HVM_PARAM_VMPORT_REGS_PFN.
    cpu_by_ioreq_id name implies the array is indexed by an id
    carries in the ioreq.
      Renamed cpu_by_ioreq_id to cpu_by_vcpu_id.
    Is cpu_get_vmport_ioreq_from_shared_memory worth its own
    function?
      Moved in-line.
    I don't think you need the barrier anyway.
      Dropped the barrier.
    Oh, I now realize you mean the same theoretical rather than
    actual limit, in which case this can be a build time check
    anyway.
      Switch to build time check, move to a better place.
    You could avoid passing state to both of them by setting
    current_cpu here couldn't you?
      Yes, moved state usage to handle_vmport_ioreq().
    Error out if it fails with error != -ENOSYS.
      Done.

v4:
    Please try to get rid of the #ifdefs.
      Moved 2 #ifdefs into hw/xen/xen_common.h


 include/hw/xen/xen_common.h |  22 +++++++++
 xen-hvm.c                   | 108 ++++++++++++++++++++++++++++++++++++++++++--
 2 files changed, 125 insertions(+), 5 deletions(-)

diff --git a/include/hw/xen/xen_common.h b/include/hw/xen/xen_common.h
index 07731b9..42e3d77 100644
--- a/include/hw/xen/xen_common.h
+++ b/include/hw/xen/xen_common.h
@@ -164,4 +164,26 @@ void destroy_hvm_domain(bool reboot);
 /* shutdown/destroy current domain because of an error */
 void xen_shutdown_fatal_error(const char *fmt, ...) GCC_FMT_ATTR(1, 2);
 
+#ifdef HVM_PARAM_VMPORT_REGS_PFN
+static inline int xen_get_vmport_regs_pfn(XenXC xc, domid_t dom,
+                                          unsigned long *vmport_regs_pfn)
+{
+    return xc_get_hvm_param(xc, dom, HVM_PARAM_VMPORT_REGS_PFN,
+                            vmport_regs_pfn);
+}
+#else
+static inline int xen_get_vmport_regs_pfn(XenXC xc, domid_t dom,
+                                          unsigned long *vmport_regs_pfn)
+{
+    return -ENOSYS;
+}
+#endif
+
+#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)
+/* Force a compilation error if condition is true */
+#define BUILD_BUG_ON(cond) ({ _Static_assert(!(cond), "!(" #cond ")"); })
+#else
+#define BUILD_BUG_ON(cond) ((void)sizeof(struct { int:-!!(cond); }))
+#endif
+
 #endif /* QEMU_HW_XEN_COMMON_H */
diff --git a/xen-hvm.c b/xen-hvm.c
index 05e522c..7c6291a 100644
--- a/xen-hvm.c
+++ b/xen-hvm.c
@@ -41,6 +41,29 @@ static MemoryRegion *framebuffer;
 static bool xen_in_migration;
 
 /* Compatibility with older version */
+
+/* This allows QEMU to build on a system that has Xen 4.5 or earlier
+ * installed.  This here (not in hw/xen/xen_common.h) because xen/hvm/ioreq.h
+ * needs to be included before this block and hw/xen/xen_common.h needs to
+ * be included before xen/hvm/ioreq.h
+ */
+#ifndef IOREQ_TYPE_VMWARE_PORT
+#define IOREQ_TYPE_VMWARE_PORT  3
+struct vmware_regs {
+    uint32_t esi;
+    uint32_t edi;
+    uint32_t ebx;
+    uint32_t ecx;
+    uint32_t edx;
+};
+typedef struct vmware_regs vmware_regs_t;
+
+struct shared_vmport_iopage {
+    struct vmware_regs vcpu_vmport_regs[1];
+};
+typedef struct shared_vmport_iopage shared_vmport_iopage_t;
+#endif
+
 #if __XEN_LATEST_INTERFACE_VERSION__ < 0x0003020a
 static inline uint32_t xen_vcpu_eport(shared_iopage_t *shared_page, int i)
 {
@@ -79,8 +102,10 @@ typedef struct XenPhysmap {
 
 typedef struct XenIOState {
     shared_iopage_t *shared_page;
+    shared_vmport_iopage_t *shared_vmport_page;
     buffered_iopage_t *buffered_io_page;
     QEMUTimer *buffered_io_timer;
+    CPUState **cpu_by_vcpu_id;
     /* the evtchn port for polling the notification, */
     evtchn_port_t *ioreq_local_port;
     /* evtchn local port for buffered io */
@@ -773,7 +798,50 @@ static void cpu_ioreq_move(ioreq_t *req)
     }
 }
 
-static void handle_ioreq(ioreq_t *req)
+static void regs_to_cpu(vmware_regs_t *vmport_regs, ioreq_t *req)
+{
+    X86CPU *cpu;
+    CPUX86State *env;
+
+    cpu = X86_CPU(current_cpu);
+    env = &cpu->env;
+    env->regs[R_EAX] = req->data;
+    env->regs[R_EBX] = vmport_regs->ebx;
+    env->regs[R_ECX] = vmport_regs->ecx;
+    env->regs[R_EDX] = vmport_regs->edx;
+    env->regs[R_ESI] = vmport_regs->esi;
+    env->regs[R_EDI] = vmport_regs->edi;
+}
+
+static void regs_from_cpu(vmware_regs_t *vmport_regs)
+{
+    X86CPU *cpu = X86_CPU(current_cpu);
+    CPUX86State *env = &cpu->env;
+
+    vmport_regs->ebx = env->regs[R_EBX];
+    vmport_regs->ecx = env->regs[R_ECX];
+    vmport_regs->edx = env->regs[R_EDX];
+    vmport_regs->esi = env->regs[R_ESI];
+    vmport_regs->edi = env->regs[R_EDI];
+}
+
+static void handle_vmport_ioreq(XenIOState *state, ioreq_t *req)
+{
+    vmware_regs_t *vmport_regs;
+
+    assert(state->shared_vmport_page);
+    vmport_regs =
+        &state->shared_vmport_page->vcpu_vmport_regs[state->send_vcpu];
+    BUILD_BUG_ON(sizeof(*req) < sizeof(*vmport_regs));
+
+    current_cpu = state->cpu_by_vcpu_id[state->send_vcpu];
+    regs_to_cpu(vmport_regs, req);
+    cpu_ioreq_pio(req);
+    regs_from_cpu(vmport_regs);
+    current_cpu = NULL;
+}
+
+static void handle_ioreq(XenIOState *state, ioreq_t *req)
 {
     if (!req->data_is_ptr && (req->dir == IOREQ_WRITE) &&
             (req->size < sizeof (target_ulong))) {
@@ -787,6 +855,9 @@ static void handle_ioreq(ioreq_t *req)
         case IOREQ_TYPE_COPY:
             cpu_ioreq_move(req);
             break;
+        case IOREQ_TYPE_VMWARE_PORT:
+            handle_vmport_ioreq(state, req);
+            break;
         case IOREQ_TYPE_TIMEOFFSET:
             break;
         case IOREQ_TYPE_INVALIDATE:
@@ -828,7 +899,7 @@ static int handle_buffered_iopage(XenIOState *state)
             req.data |= ((uint64_t)buf_req->data) << 32;
         }
 
-        handle_ioreq(&req);
+        handle_ioreq(state, &req);
 
         xen_mb();
         state->buffered_io_page->read_pointer += qw ? 2 : 1;
@@ -857,14 +928,16 @@ static void cpu_handle_ioreq(void *opaque)
 
     handle_buffered_iopage(state);
     if (req) {
-        handle_ioreq(req);
+        handle_ioreq(state, req);
 
         if (req->state != STATE_IOREQ_INPROCESS) {
             fprintf(stderr, "Badness in I/O request ... not in service?!: "
                     "%x, ptr: %x, port: %"PRIx64", "
-                    "data: %"PRIx64", count: %" FMT_ioreq_size ", size: %" FMT_ioreq_size "\n",
+                    "data: %"PRIx64", count: %" FMT_ioreq_size
+                    ", size: %" FMT_ioreq_size
+                    ", type: %"FMT_ioreq_size"\n",
                     req->state, req->data_is_ptr, req->addr,
-                    req->data, req->count, req->size);
+                    req->data, req->count, req->size, req->type);
             destroy_hvm_domain(false);
             return;
         }
@@ -904,6 +977,14 @@ static void xen_main_loop_prepare(XenIOState *state)
                                                  state);
 
     if (evtchn_fd != -1) {
+        CPUState *cpu_state;
+
+        DPRINTF("%s: Init cpu_by_vcpu_id\n", __func__);
+        CPU_FOREACH(cpu_state) {
+            DPRINTF("%s: cpu_by_vcpu_id[%d]=%p\n",
+                    __func__, cpu_state->cpu_index, cpu_state);
+            state->cpu_by_vcpu_id[cpu_state->cpu_index] = cpu_state;
+        }
         qemu_set_fd_handler(evtchn_fd, cpu_handle_ioreq, NULL, state);
     }
 }
@@ -1020,6 +1101,20 @@ int xen_hvm_init(ram_addr_t *below_4g_mem_size, ram_addr_t *above_4g_mem_size,
                  errno, xen_xc);
     }
 
+    rc = xen_get_vmport_regs_pfn(xen_xc, xen_domid, &ioreq_pfn);
+    if (!rc) {
+        DPRINTF("shared vmport page at pfn %lx\n", ioreq_pfn);
+        state->shared_vmport_page =
+            xc_map_foreign_range(xen_xc, xen_domid, XC_PAGE_SIZE,
+                                 PROT_READ|PROT_WRITE, ioreq_pfn);
+        if (state->shared_vmport_page == NULL) {
+            hw_error("map shared vmport IO page returned error %d handle="
+                     XC_INTERFACE_FMT, errno, xen_xc);
+        }
+    } else if (rc != -ENOSYS) {
+        hw_error("get vmport regs pfn returned error %d, rc=%d", errno, rc);
+    }
+
     xc_get_hvm_param(xen_xc, xen_domid, HVM_PARAM_BUFIOREQ_PFN, &ioreq_pfn);
     DPRINTF("buffered io page at pfn %lx\n", ioreq_pfn);
     state->buffered_io_page = xc_map_foreign_range(xen_xc, xen_domid, XC_PAGE_SIZE,
@@ -1028,6 +1123,9 @@ int xen_hvm_init(ram_addr_t *below_4g_mem_size, ram_addr_t *above_4g_mem_size,
         hw_error("map buffered IO page returned error %d", errno);
     }
 
+    /* Note: cpus is empty at this point in init */
+    state->cpu_by_vcpu_id = g_malloc0(max_cpus * sizeof(CPUState *));
+
     state->ioreq_local_port = g_malloc0(max_cpus * sizeof (evtchn_port_t));
 
     /* FIXME: how about if we overflow the page here? */
-- 
1.8.4

Re: [Qemu-devel] [PATCH v6 1/1] xen-hvm.c: Add support for Xen access to vmport

[Stefano Stabellini]

On Mon, 20 Oct 2014, Don Slutz wrote:
> This adds synchronisation of the 6 vcpu registers (only 32bits of
> them) that vmport.c needs between Xen and QEMU.
> 
> This is to avoid a 2nd and 3rd exchange between QEMU and Xen to
> fetch and put these 6 vcpu registers used by the code in vmport.c
> and vmmouse.c
> 
> The registers are passed in the new shared page provided by
> HVM_PARAM_VMPORT_REGS_PFN.
> 
> Add new array to XenIOState that allows selection of current_cpu by
> vcpu id.
> 
> Now pass XenIOState to handle_ioreq().
> 
> Add new routines regs_to_cpu(), regs_from_cpu(), and
> handle_vmport_ioreq().
> 
> Signed-off-by: Don Slutz <dslutz@verizon.com>
> Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
> ---
> v6:
>     Do we need a forward declaration?
>       Nope. Drooped.
>     Added Reviewed-by from Paul Durrant.
> 
> v5:
>     vmware_ioreq_t struct is not really a request any more. Maybe
>     vmware_regs_t?
>       Renamed various parts from vmware_ioreq to vmware_regs.  Also
>       HVM_PARAM_VMPORT_IOREQ_PFN to HVM_PARAM_VMPORT_REGS_PFN.
>     cpu_by_ioreq_id name implies the array is indexed by an id
>     carries in the ioreq.
>       Renamed cpu_by_ioreq_id to cpu_by_vcpu_id.
>     Is cpu_get_vmport_ioreq_from_shared_memory worth its own
>     function?
>       Moved in-line.
>     I don't think you need the barrier anyway.
>       Dropped the barrier.
>     Oh, I now realize you mean the same theoretical rather than
>     actual limit, in which case this can be a build time check
>     anyway.
>       Switch to build time check, move to a better place.
>     You could avoid passing state to both of them by setting
>     current_cpu here couldn't you?
>       Yes, moved state usage to handle_vmport_ioreq().
>     Error out if it fails with error != -ENOSYS.
>       Done.
> 
> v4:
>     Please try to get rid of the #ifdefs.
>       Moved 2 #ifdefs into hw/xen/xen_common.h
> 
> 
>  include/hw/xen/xen_common.h |  22 +++++++++
>  xen-hvm.c                   | 108 ++++++++++++++++++++++++++++++++++++++++++--
>  2 files changed, 125 insertions(+), 5 deletions(-)
> 
> diff --git a/include/hw/xen/xen_common.h b/include/hw/xen/xen_common.h
> index 07731b9..42e3d77 100644
> --- a/include/hw/xen/xen_common.h
> +++ b/include/hw/xen/xen_common.h
> @@ -164,4 +164,26 @@ void destroy_hvm_domain(bool reboot);
>  /* shutdown/destroy current domain because of an error */
>  void xen_shutdown_fatal_error(const char *fmt, ...) GCC_FMT_ATTR(1, 2);
>  
> +#ifdef HVM_PARAM_VMPORT_REGS_PFN
> +static inline int xen_get_vmport_regs_pfn(XenXC xc, domid_t dom,
> +                                          unsigned long *vmport_regs_pfn)
> +{
> +    return xc_get_hvm_param(xc, dom, HVM_PARAM_VMPORT_REGS_PFN,
> +                            vmport_regs_pfn);
> +}
> +#else
> +static inline int xen_get_vmport_regs_pfn(XenXC xc, domid_t dom,
> +                                          unsigned long *vmport_regs_pfn)
> +{
> +    return -ENOSYS;
> +}
> +#endif
> +
> +#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)
> +/* Force a compilation error if condition is true */
> +#define BUILD_BUG_ON(cond) ({ _Static_assert(!(cond), "!(" #cond ")"); })
> +#else
> +#define BUILD_BUG_ON(cond) ((void)sizeof(struct { int:-!!(cond); }))
> +#endif

Use QEMU_BUILD_BUG_ON, please do not define your own. But aside from
this, it is OK for me.


>  #endif /* QEMU_HW_XEN_COMMON_H */
> diff --git a/xen-hvm.c b/xen-hvm.c
> index 05e522c..7c6291a 100644
> --- a/xen-hvm.c
> +++ b/xen-hvm.c
> @@ -41,6 +41,29 @@ static MemoryRegion *framebuffer;
>  static bool xen_in_migration;
>  
>  /* Compatibility with older version */
> +
> +/* This allows QEMU to build on a system that has Xen 4.5 or earlier
> + * installed.  This here (not in hw/xen/xen_common.h) because xen/hvm/ioreq.h
> + * needs to be included before this block and hw/xen/xen_common.h needs to
> + * be included before xen/hvm/ioreq.h
> + */
> +#ifndef IOREQ_TYPE_VMWARE_PORT
> +#define IOREQ_TYPE_VMWARE_PORT  3
> +struct vmware_regs {
> +    uint32_t esi;
> +    uint32_t edi;
> +    uint32_t ebx;
> +    uint32_t ecx;
> +    uint32_t edx;
> +};
> +typedef struct vmware_regs vmware_regs_t;
> +
> +struct shared_vmport_iopage {
> +    struct vmware_regs vcpu_vmport_regs[1];
> +};
> +typedef struct shared_vmport_iopage shared_vmport_iopage_t;
> +#endif
> +
>  #if __XEN_LATEST_INTERFACE_VERSION__ < 0x0003020a
>  static inline uint32_t xen_vcpu_eport(shared_iopage_t *shared_page, int i)
>  {
> @@ -79,8 +102,10 @@ typedef struct XenPhysmap {
>  
>  typedef struct XenIOState {
>      shared_iopage_t *shared_page;
> +    shared_vmport_iopage_t *shared_vmport_page;
>      buffered_iopage_t *buffered_io_page;
>      QEMUTimer *buffered_io_timer;
> +    CPUState **cpu_by_vcpu_id;
>      /* the evtchn port for polling the notification, */
>      evtchn_port_t *ioreq_local_port;
>      /* evtchn local port for buffered io */
> @@ -773,7 +798,50 @@ static void cpu_ioreq_move(ioreq_t *req)
>      }
>  }
>  
> -static void handle_ioreq(ioreq_t *req)
> +static void regs_to_cpu(vmware_regs_t *vmport_regs, ioreq_t *req)
> +{
> +    X86CPU *cpu;
> +    CPUX86State *env;
> +
> +    cpu = X86_CPU(current_cpu);
> +    env = &cpu->env;
> +    env->regs[R_EAX] = req->data;
> +    env->regs[R_EBX] = vmport_regs->ebx;
> +    env->regs[R_ECX] = vmport_regs->ecx;
> +    env->regs[R_EDX] = vmport_regs->edx;
> +    env->regs[R_ESI] = vmport_regs->esi;
> +    env->regs[R_EDI] = vmport_regs->edi;
> +}
> +
> +static void regs_from_cpu(vmware_regs_t *vmport_regs)
> +{
> +    X86CPU *cpu = X86_CPU(current_cpu);
> +    CPUX86State *env = &cpu->env;
> +
> +    vmport_regs->ebx = env->regs[R_EBX];
> +    vmport_regs->ecx = env->regs[R_ECX];
> +    vmport_regs->edx = env->regs[R_EDX];
> +    vmport_regs->esi = env->regs[R_ESI];
> +    vmport_regs->edi = env->regs[R_EDI];
> +}
> +
> +static void handle_vmport_ioreq(XenIOState *state, ioreq_t *req)
> +{
> +    vmware_regs_t *vmport_regs;
> +
> +    assert(state->shared_vmport_page);
> +    vmport_regs =
> +        &state->shared_vmport_page->vcpu_vmport_regs[state->send_vcpu];
> +    BUILD_BUG_ON(sizeof(*req) < sizeof(*vmport_regs));

QEMU_BUILD_BUG_ON


> +    current_cpu = state->cpu_by_vcpu_id[state->send_vcpu];
> +    regs_to_cpu(vmport_regs, req);
> +    cpu_ioreq_pio(req);
> +    regs_from_cpu(vmport_regs);
> +    current_cpu = NULL;
> +}
> +
> +static void handle_ioreq(XenIOState *state, ioreq_t *req)
>  {
>      if (!req->data_is_ptr && (req->dir == IOREQ_WRITE) &&
>              (req->size < sizeof (target_ulong))) {
> @@ -787,6 +855,9 @@ static void handle_ioreq(ioreq_t *req)
>          case IOREQ_TYPE_COPY:
>              cpu_ioreq_move(req);
>              break;
> +        case IOREQ_TYPE_VMWARE_PORT:
> +            handle_vmport_ioreq(state, req);
> +            break;
>          case IOREQ_TYPE_TIMEOFFSET:
>              break;
>          case IOREQ_TYPE_INVALIDATE:
> @@ -828,7 +899,7 @@ static int handle_buffered_iopage(XenIOState *state)
>              req.data |= ((uint64_t)buf_req->data) << 32;
>          }
>  
> -        handle_ioreq(&req);
> +        handle_ioreq(state, &req);
>  
>          xen_mb();
>          state->buffered_io_page->read_pointer += qw ? 2 : 1;
> @@ -857,14 +928,16 @@ static void cpu_handle_ioreq(void *opaque)
>  
>      handle_buffered_iopage(state);
>      if (req) {
> -        handle_ioreq(req);
> +        handle_ioreq(state, req);
>  
>          if (req->state != STATE_IOREQ_INPROCESS) {
>              fprintf(stderr, "Badness in I/O request ... not in service?!: "
>                      "%x, ptr: %x, port: %"PRIx64", "
> -                    "data: %"PRIx64", count: %" FMT_ioreq_size ", size: %" FMT_ioreq_size "\n",
> +                    "data: %"PRIx64", count: %" FMT_ioreq_size
> +                    ", size: %" FMT_ioreq_size
> +                    ", type: %"FMT_ioreq_size"\n",
>                      req->state, req->data_is_ptr, req->addr,
> -                    req->data, req->count, req->size);
> +                    req->data, req->count, req->size, req->type);
>              destroy_hvm_domain(false);
>              return;
>          }
> @@ -904,6 +977,14 @@ static void xen_main_loop_prepare(XenIOState *state)
>                                                   state);
>  
>      if (evtchn_fd != -1) {
> +        CPUState *cpu_state;
> +
> +        DPRINTF("%s: Init cpu_by_vcpu_id\n", __func__);
> +        CPU_FOREACH(cpu_state) {
> +            DPRINTF("%s: cpu_by_vcpu_id[%d]=%p\n",
> +                    __func__, cpu_state->cpu_index, cpu_state);
> +            state->cpu_by_vcpu_id[cpu_state->cpu_index] = cpu_state;
> +        }
>          qemu_set_fd_handler(evtchn_fd, cpu_handle_ioreq, NULL, state);
>      }
>  }
> @@ -1020,6 +1101,20 @@ int xen_hvm_init(ram_addr_t *below_4g_mem_size, ram_addr_t *above_4g_mem_size,
>                   errno, xen_xc);
>      }
>  
> +    rc = xen_get_vmport_regs_pfn(xen_xc, xen_domid, &ioreq_pfn);
> +    if (!rc) {
> +        DPRINTF("shared vmport page at pfn %lx\n", ioreq_pfn);
> +        state->shared_vmport_page =
> +            xc_map_foreign_range(xen_xc, xen_domid, XC_PAGE_SIZE,
> +                                 PROT_READ|PROT_WRITE, ioreq_pfn);
> +        if (state->shared_vmport_page == NULL) {
> +            hw_error("map shared vmport IO page returned error %d handle="
> +                     XC_INTERFACE_FMT, errno, xen_xc);
> +        }
> +    } else if (rc != -ENOSYS) {
> +        hw_error("get vmport regs pfn returned error %d, rc=%d", errno, rc);
> +    }
> +
>      xc_get_hvm_param(xen_xc, xen_domid, HVM_PARAM_BUFIOREQ_PFN, &ioreq_pfn);
>      DPRINTF("buffered io page at pfn %lx\n", ioreq_pfn);
>      state->buffered_io_page = xc_map_foreign_range(xen_xc, xen_domid, XC_PAGE_SIZE,
> @@ -1028,6 +1123,9 @@ int xen_hvm_init(ram_addr_t *below_4g_mem_size, ram_addr_t *above_4g_mem_size,
>          hw_error("map buffered IO page returned error %d", errno);
>      }
>  
> +    /* Note: cpus is empty at this point in init */
> +    state->cpu_by_vcpu_id = g_malloc0(max_cpus * sizeof(CPUState *));
> +
>      state->ioreq_local_port = g_malloc0(max_cpus * sizeof (evtchn_port_t));
>  
>      /* FIXME: how about if we overflow the page here? */
> -- 
> 1.8.4
> 

Re: [Qemu-devel] [PATCH v6 1/1] xen-hvm.c: Add support for Xen access to vmport

[Don Slutz]

On 10/20/14 11:03, Stefano Stabellini wrote:
> On Mon, 20 Oct 2014, Don Slutz wrote:
>> This adds synchronisation of the 6 vcpu registers (only 32bits of
>> them) that vmport.c needs between Xen and QEMU.
>>
>> This is to avoid a 2nd and 3rd exchange between QEMU and Xen to
>> fetch and put these 6 vcpu registers used by the code in vmport.c
>> and vmmouse.c
>>
>> The registers are passed in the new shared page provided by
>> HVM_PARAM_VMPORT_REGS_PFN.
>>
>> Add new array to XenIOState that allows selection of current_cpu by
>> vcpu id.
>>
>> Now pass XenIOState to handle_ioreq().
>>
>> Add new routines regs_to_cpu(), regs_from_cpu(), and
>> handle_vmport_ioreq().
>>
>> Signed-off-by: Don Slutz <dslutz@verizon.com>
>> Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
>> ---
>> v6:
>>      Do we need a forward declaration?
>>        Nope. Drooped.
>>      Added Reviewed-by from Paul Durrant.
>>
>> v5:
>>      vmware_ioreq_t struct is not really a request any more. Maybe
>>      vmware_regs_t?
>>        Renamed various parts from vmware_ioreq to vmware_regs.  Also
>>        HVM_PARAM_VMPORT_IOREQ_PFN to HVM_PARAM_VMPORT_REGS_PFN.
>>      cpu_by_ioreq_id name implies the array is indexed by an id
>>      carries in the ioreq.
>>        Renamed cpu_by_ioreq_id to cpu_by_vcpu_id.
>>      Is cpu_get_vmport_ioreq_from_shared_memory worth its own
>>      function?
>>        Moved in-line.
>>      I don't think you need the barrier anyway.
>>        Dropped the barrier.
>>      Oh, I now realize you mean the same theoretical rather than
>>      actual limit, in which case this can be a build time check
>>      anyway.
>>        Switch to build time check, move to a better place.
>>      You could avoid passing state to both of them by setting
>>      current_cpu here couldn't you?
>>        Yes, moved state usage to handle_vmport_ioreq().
>>      Error out if it fails with error != -ENOSYS.
>>        Done.
>>
>> v4:
>>      Please try to get rid of the #ifdefs.
>>        Moved 2 #ifdefs into hw/xen/xen_common.h
>>
>>
>>   include/hw/xen/xen_common.h |  22 +++++++++
>>   xen-hvm.c                   | 108 ++++++++++++++++++++++++++++++++++++++++++--
>>   2 files changed, 125 insertions(+), 5 deletions(-)
>>
>> diff --git a/include/hw/xen/xen_common.h b/include/hw/xen/xen_common.h
>> index 07731b9..42e3d77 100644
>> --- a/include/hw/xen/xen_common.h
>> +++ b/include/hw/xen/xen_common.h
>> @@ -164,4 +164,26 @@ void destroy_hvm_domain(bool reboot);
>>   /* shutdown/destroy current domain because of an error */
>>   void xen_shutdown_fatal_error(const char *fmt, ...) GCC_FMT_ATTR(1, 2);
>>   
>> +#ifdef HVM_PARAM_VMPORT_REGS_PFN
>> +static inline int xen_get_vmport_regs_pfn(XenXC xc, domid_t dom,
>> +                                          unsigned long *vmport_regs_pfn)
>> +{
>> +    return xc_get_hvm_param(xc, dom, HVM_PARAM_VMPORT_REGS_PFN,
>> +                            vmport_regs_pfn);
>> +}
>> +#else
>> +static inline int xen_get_vmport_regs_pfn(XenXC xc, domid_t dom,
>> +                                          unsigned long *vmport_regs_pfn)
>> +{
>> +    return -ENOSYS;
>> +}
>> +#endif
>> +
>> +#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)
>> +/* Force a compilation error if condition is true */
>> +#define BUILD_BUG_ON(cond) ({ _Static_assert(!(cond), "!(" #cond ")"); })
>> +#else
>> +#define BUILD_BUG_ON(cond) ((void)sizeof(struct { int:-!!(cond); }))
>> +#endif
> Use QEMU_BUILD_BUG_ON, please do not define your own. But aside from
> this, it is OK for me.
>

Sigh, I looked for this and failed to find it.  Will switch.
v7 soon.

     -Don Slutz

>>   #endif /* QEMU_HW_XEN_COMMON_H */
>> diff --git a/xen-hvm.c b/xen-hvm.c
>> index 05e522c..7c6291a 100644
>> --- a/xen-hvm.c
>> +++ b/xen-hvm.c
>> @@ -41,6 +41,29 @@ static MemoryRegion *framebuffer;
>>   static bool xen_in_migration;
>>   
>>   /* Compatibility with older version */
>> +
>> +/* This allows QEMU to build on a system that has Xen 4.5 or earlier
>> + * installed.  This here (not in hw/xen/xen_common.h) because xen/hvm/ioreq.h
>> + * needs to be included before this block and hw/xen/xen_common.h needs to
>> + * be included before xen/hvm/ioreq.h
>> + */
>> +#ifndef IOREQ_TYPE_VMWARE_PORT
>> +#define IOREQ_TYPE_VMWARE_PORT  3
>> +struct vmware_regs {
>> +    uint32_t esi;
>> +    uint32_t edi;
>> +    uint32_t ebx;
>> +    uint32_t ecx;
>> +    uint32_t edx;
>> +};
>> +typedef struct vmware_regs vmware_regs_t;
>> +
>> +struct shared_vmport_iopage {
>> +    struct vmware_regs vcpu_vmport_regs[1];
>> +};
>> +typedef struct shared_vmport_iopage shared_vmport_iopage_t;
>> +#endif
>> +
>>   #if __XEN_LATEST_INTERFACE_VERSION__ < 0x0003020a
>>   static inline uint32_t xen_vcpu_eport(shared_iopage_t *shared_page, int i)
>>   {
>> @@ -79,8 +102,10 @@ typedef struct XenPhysmap {
>>   
>>   typedef struct XenIOState {
>>       shared_iopage_t *shared_page;
>> +    shared_vmport_iopage_t *shared_vmport_page;
>>       buffered_iopage_t *buffered_io_page;
>>       QEMUTimer *buffered_io_timer;
>> +    CPUState **cpu_by_vcpu_id;
>>       /* the evtchn port for polling the notification, */
>>       evtchn_port_t *ioreq_local_port;
>>       /* evtchn local port for buffered io */
>> @@ -773,7 +798,50 @@ static void cpu_ioreq_move(ioreq_t *req)
>>       }
>>   }
>>   
>> -static void handle_ioreq(ioreq_t *req)
>> +static void regs_to_cpu(vmware_regs_t *vmport_regs, ioreq_t *req)
>> +{
>> +    X86CPU *cpu;
>> +    CPUX86State *env;
>> +
>> +    cpu = X86_CPU(current_cpu);
>> +    env = &cpu->env;
>> +    env->regs[R_EAX] = req->data;
>> +    env->regs[R_EBX] = vmport_regs->ebx;
>> +    env->regs[R_ECX] = vmport_regs->ecx;
>> +    env->regs[R_EDX] = vmport_regs->edx;
>> +    env->regs[R_ESI] = vmport_regs->esi;
>> +    env->regs[R_EDI] = vmport_regs->edi;
>> +}
>> +
>> +static void regs_from_cpu(vmware_regs_t *vmport_regs)
>> +{
>> +    X86CPU *cpu = X86_CPU(current_cpu);
>> +    CPUX86State *env = &cpu->env;
>> +
>> +    vmport_regs->ebx = env->regs[R_EBX];
>> +    vmport_regs->ecx = env->regs[R_ECX];
>> +    vmport_regs->edx = env->regs[R_EDX];
>> +    vmport_regs->esi = env->regs[R_ESI];
>> +    vmport_regs->edi = env->regs[R_EDI];
>> +}
>> +
>> +static void handle_vmport_ioreq(XenIOState *state, ioreq_t *req)
>> +{
>> +    vmware_regs_t *vmport_regs;
>> +
>> +    assert(state->shared_vmport_page);
>> +    vmport_regs =
>> +        &state->shared_vmport_page->vcpu_vmport_regs[state->send_vcpu];
>> +    BUILD_BUG_ON(sizeof(*req) < sizeof(*vmport_regs));
> QEMU_BUILD_BUG_ON
>
>
>> +    current_cpu = state->cpu_by_vcpu_id[state->send_vcpu];
>> +    regs_to_cpu(vmport_regs, req);
>> +    cpu_ioreq_pio(req);
>> +    regs_from_cpu(vmport_regs);
>> +    current_cpu = NULL;
>> +}
>> +
>> +static void handle_ioreq(XenIOState *state, ioreq_t *req)
>>   {
>>       if (!req->data_is_ptr && (req->dir == IOREQ_WRITE) &&
>>               (req->size < sizeof (target_ulong))) {
>> @@ -787,6 +855,9 @@ static void handle_ioreq(ioreq_t *req)
>>           case IOREQ_TYPE_COPY:
>>               cpu_ioreq_move(req);
>>               break;
>> +        case IOREQ_TYPE_VMWARE_PORT:
>> +            handle_vmport_ioreq(state, req);
>> +            break;
>>           case IOREQ_TYPE_TIMEOFFSET:
>>               break;
>>           case IOREQ_TYPE_INVALIDATE:
>> @@ -828,7 +899,7 @@ static int handle_buffered_iopage(XenIOState *state)
>>               req.data |= ((uint64_t)buf_req->data) << 32;
>>           }
>>   
>> -        handle_ioreq(&req);
>> +        handle_ioreq(state, &req);
>>   
>>           xen_mb();
>>           state->buffered_io_page->read_pointer += qw ? 2 : 1;
>> @@ -857,14 +928,16 @@ static void cpu_handle_ioreq(void *opaque)
>>   
>>       handle_buffered_iopage(state);
>>       if (req) {
>> -        handle_ioreq(req);
>> +        handle_ioreq(state, req);
>>   
>>           if (req->state != STATE_IOREQ_INPROCESS) {
>>               fprintf(stderr, "Badness in I/O request ... not in service?!: "
>>                       "%x, ptr: %x, port: %"PRIx64", "
>> -                    "data: %"PRIx64", count: %" FMT_ioreq_size ", size: %" FMT_ioreq_size "\n",
>> +                    "data: %"PRIx64", count: %" FMT_ioreq_size
>> +                    ", size: %" FMT_ioreq_size
>> +                    ", type: %"FMT_ioreq_size"\n",
>>                       req->state, req->data_is_ptr, req->addr,
>> -                    req->data, req->count, req->size);
>> +                    req->data, req->count, req->size, req->type);
>>               destroy_hvm_domain(false);
>>               return;
>>           }
>> @@ -904,6 +977,14 @@ static void xen_main_loop_prepare(XenIOState *state)
>>                                                    state);
>>   
>>       if (evtchn_fd != -1) {
>> +        CPUState *cpu_state;
>> +
>> +        DPRINTF("%s: Init cpu_by_vcpu_id\n", __func__);
>> +        CPU_FOREACH(cpu_state) {
>> +            DPRINTF("%s: cpu_by_vcpu_id[%d]=%p\n",
>> +                    __func__, cpu_state->cpu_index, cpu_state);
>> +            state->cpu_by_vcpu_id[cpu_state->cpu_index] = cpu_state;
>> +        }
>>           qemu_set_fd_handler(evtchn_fd, cpu_handle_ioreq, NULL, state);
>>       }
>>   }
>> @@ -1020,6 +1101,20 @@ int xen_hvm_init(ram_addr_t *below_4g_mem_size, ram_addr_t *above_4g_mem_size,
>>                    errno, xen_xc);
>>       }
>>   
>> +    rc = xen_get_vmport_regs_pfn(xen_xc, xen_domid, &ioreq_pfn);
>> +    if (!rc) {
>> +        DPRINTF("shared vmport page at pfn %lx\n", ioreq_pfn);
>> +        state->shared_vmport_page =
>> +            xc_map_foreign_range(xen_xc, xen_domid, XC_PAGE_SIZE,
>> +                                 PROT_READ|PROT_WRITE, ioreq_pfn);
>> +        if (state->shared_vmport_page == NULL) {
>> +            hw_error("map shared vmport IO page returned error %d handle="
>> +                     XC_INTERFACE_FMT, errno, xen_xc);
>> +        }
>> +    } else if (rc != -ENOSYS) {
>> +        hw_error("get vmport regs pfn returned error %d, rc=%d", errno, rc);
>> +    }
>> +
>>       xc_get_hvm_param(xen_xc, xen_domid, HVM_PARAM_BUFIOREQ_PFN, &ioreq_pfn);
>>       DPRINTF("buffered io page at pfn %lx\n", ioreq_pfn);
>>       state->buffered_io_page = xc_map_foreign_range(xen_xc, xen_domid, XC_PAGE_SIZE,
>> @@ -1028,6 +1123,9 @@ int xen_hvm_init(ram_addr_t *below_4g_mem_size, ram_addr_t *above_4g_mem_size,
>>           hw_error("map buffered IO page returned error %d", errno);
>>       }
>>   
>> +    /* Note: cpus is empty at this point in init */
>> +    state->cpu_by_vcpu_id = g_malloc0(max_cpus * sizeof(CPUState *));
>> +
>>       state->ioreq_local_port = g_malloc0(max_cpus * sizeof (evtchn_port_t));
>>   
>>       /* FIXME: how about if we overflow the page here? */
>> -- 
>> 1.8.4
>>